Harvest-Time Protein Shocks
and Price Adjustment in U.S.
Wheat Markets
Barry K. Goodwin and Vincent H. Smith*
Agricultural Marketing Policy Paper No. 5
Objective Analysis
June 2005
for Informed
Decision Making
*
Barry Goodwin is professor of agricultural economics at North Carolina State University. Vince
Smith is professor of agricultural economics at Montana State University and co-director of the
Agricultural Marketing Policy Center.
Abstract
Dynamic relationships between three classes of wheat are investigated using threshold VAR models
incorporating the effects of protein availability. Changes in the stock of protein are found to generate
significant impulse responses in the price of hard red spring wheat and hard red winter wheat but not soft red
wheat. These impulse responses to identical changes in protein stocks are larger when the absolute deviations
of protein stocks from normal levels are large. Shocks to the prices of individual classes of wheat result in
complex impulse responses in the prices of the other wheats. Notably, however, a shock to the price of hard red
winter weak appears to result in little or no impulse response in the price of hard spring wheat, though,
importantly, the opposite is not true.
Agricultural commodities such as wheat are
typically heterogeneous, with quality characteristics
that differ across space, time, and variety. The
extent to which market prices account for such
quality differences has been an important issue to
the overall efficiency of markets for agricultural
commodities. The benefits associated with accurate
measurement of qualities by buyers and sellers in a
market must be weighed against the potential costs
associated with such an accurate quality assessment.
Some characteristics (foreign matter, shrunken and
broken kernels, etc.) are easy to measure while
others (valorimeter and farinograph measures) are
much more difficult and expensive to identify.
Protein content is one of the most basic quality
characteristics shaping the potential utility of a
particular class of wheat for various uses. It plays
such an important role in price interrelationships
among different types and grades of wheat that it
also forms the basis for U.S. standard variety
grades. For example, higher protein wheat
varieties such as dark northern spring and hard red
winter typically command a price premium over
wheat varieties with lower protein contents (for
example, see Espinosa and Goodwin, 1991), and
that the price premium varies over time almost
surely in accord with shifts in supply and demand
for that attribute (Parcell and Stiegert, 1998), as
implied by the theoretical hedonic pricing
framework developed by Rosen (1974).
Several studies have examined the dynamics of
domestic and international wheat price relationships
(see, for example, Goodwin and Schroeder, 1991;
International Trade Commission, 1994; Mohanty,
Meyers, and Smith,1999). However, relatively little
attention has been directed toward interrelationships
among different types of wheat prices and quality
shocks that may relate to the aggregate level of
quality. Failure to account for these shocks is likely
to distort estimates of these relationships and
provide misleading assessments of the extent to
which prices of different types of wheat are related
to one another and, particularly, the extent to which
different types of wheat are substitutes for one
another. 1
In this paper, we are primarily concerned with the
aggregate market for protein (wheat gluten) and its
effect of price relationships among different classes
of wheat. We consider multivariate time-series
models for three classes of wheat—hard red spring,
hard red winter, and soft red winter. We are
interested in quantifying the relationships between
the protein content associated with each year’s
harvest for each type of wheat and the differentials
(reflecting protein supply and demand effects)
between various classes of wheat. Monthly price
data are used in conjunction new data constructed
by the authors that report the average (aggregate)
protein content associated with each year’s U.S.
harvest.
____________________________
1
The issue of elasticities of substitution among wheat classes
has been addressed by two recent studies by Marsh and Barnes
and Shields using structural models of derived demand
estimated with annual data. Both studies, while providing
different estimates, find that wheat is not just wheat, in the
sense that elasticities of substitution among different classes of
wheat are by no means as large as has been suggested by some
researchers (for example, Alston, Sumner and Gray).
The relationships between wheat prices and protein
content may vary substantially from year-to-year,
depending on overall wheat yields and other quality
factors.2 Further, protein content in any given year
may be affected by the characteristics of the market
for protein in preceding years, since grain stocks are
held from year to year and production practices and
variety choices may be important considerations in
the realized protein content of a wheat crop.
We use nonstructural time-series models that also
allow for costly adjustment by incorporating
threshold procedures to evaluate the effects of
protein content shocks on the time paths of wheat
prices. We account for protein availability effects
on price interrelationships from year to year and
quantify the extent to which shocks in the levels of
protein in a particular type of wheat affect the
differentials in wheat prices among the individual
wheat classes. Our analysis uses dynamic impulse
responses to track price responses to shocks in the
protein market and other shocks to specific wheat
class prices.
The analysis provides new insights about the
substitutability of different classes of wheat among
end uses, a critical issue in recent trade dispute
cases. For example, if hard red spring wheat and
hard red winter wheat are perfect or very close
substitutes, as suggested by Canadian Wheat Board
expert witnesses in testimony before the
International Trade Commission on behalf of the
Canadian Wheat Board in September, 2003, then
hard red winter prices are likely to respond rapidly
in similar ways to a shock in hard red spring prices,
and vice versa. This does not appear to be the case.
While shocks to hard red spring wheat prices
generate substantial responses in hard red winter
wheat prices, the opposite is not the case. Shocks to
hard red winter prices generate relatively weak
responses in hard red spring prices, suggesting that
hard red winter wheat is an imperfect substitute for
hard red winter wheat.
Empirical Methods
The primary objective of the empirical analysis is to
evaluate the extent to which dynamic relationships
among prices for different classes of wheat are
affected by shocks to the quality of the overall U.S.
wheat harvest. In particular, we are interested in the
role played by protein content—one of the major
determinants of the quality and functionality of
different classes and grades of wheat for different
uses. Certainly a wide variety of wheat
characteristics may be pertinent to the quality of any
given quantity of wheat. These include factors such
as variometer and farinograph measures, foreign
materials, falling numbers, ash content and so
forth.3 However, in terms of the aggregate wheat
market and the price relationships between different
types of wheats, both the results of several hedonic
studies and current industry pricing practices
indicate that each wheat harvest’s protein content is
likely to be the most relevant factor influencing
dynamic relationships among the prices of different
types of wheat.
In the spirit of the relatively extensive literature that
has addressed these issues, we adopt a standard
vector autoregression (VAR) model that includes
prices of the three major wheats—Dark Northern
Spring (DNS) in Minneapolis, Hard Red Winter
(HRW) in Kansas City, and Soft Red Winter (SRW)
in Chicago. DNS and HRW wheats typically have
much higher protein contents than SRW and are
directed toward end-uses that require stronger
gluten content (e.g., breads). We also include a
measure of the overall protein content implicit in
stocks at any point in time. Our specific measure of
this protein content variable is described in the
following section.
____________________________
2 Parcell
The paper is organized as follows. Empirical
methods are discussed in the next section. The data
are then described, empirical results are presented
and discussed, and a summary and conclusion are
provided.
and Steigert (1998) and Stiegert and Blanc
both report that the effect of a marginal increase in
protein on protein premiums varies among different
classes of wheat such as hard red spring, hard red
winter and soft red winter.
3
See Espinosa and Goodwin for a detailed discussion of how
different quality factors are related to wheat prices.
A standard VAR model can be written as:
yt = ΓX t + et ,
where yt is a vector of endogenous variables for
which dynamic adjustment paths are to be
evaluated, Γ is a matrix of parameters to be
estimated, et is a vector of random error terms, and
X t = [1, yt −1 ,..., yt − j , xt ] where xt is a vector of other
exogenous factors.
In addition to estimating a simple VAR model, we
are interested in considering the potential for
nonlinearities in the underlying relationships
represented by the VAR model. To this end, we
appeal to recent developments in the time series
literature that consider nonlinearities in the
relationships inherent in nonstructural VAR type
models. We hypothesize that adjustments to shocks
in the inherent qualities of wheat by end-users (e.g.,
bakers, millers, and food processors) are costly. In
particular, most production processes are tightly
calibrated and have specific quality requirements.
End-users may be able to make adjustments in
production processes, though these adjustments are
likely require significant technological
modifications and to be costly.4
To capture these effects, we utilize a threshold
modification to the standard VAR modeling
framework. In particular, we allow the underlying
structure of the model (represented by the
nonstructural, reduced-form parameters of the VAR
system of equations) to vary according to implied
protein availability in the market. In particular, we
consider a threshold defined by deviations from
normal levels of protein in the market. The
“normal” level of protein is defined by using a
regression of protein availability on a third-order
fourier series expansion, which is intended to
capture the large degree of seasonality that
accompanies the wheat harvests and subsequent
adjustments to stocks.
We define the “normal” level of protein (given by a
function f(t) consisting of a fourier series
expansion) by pˆ = f (t ) . Departures from normal
levels are therefore determined as:
Pt - f(t) = vt
The explicit definition of the threshold is given by
c, where the switch in regimes is triggered when
departures from normal levels of protein exceed c in
absolute value. In other words, two alternative
regimes are defined by the absolute value of vt .
The regime switching model is thus given by:
(1)
Γ
X t if

yt =  (2)
 Γ X t if
| ν t |≤ c
| ν t |> c
,
where Γ( i ) represents the parameter estimates
associated with the ith regime and c is the initially
unknown threshold parameter, which has to be
estimated. An alternative representation of this
model is as follows:
yt = (1 − δ )Γ (1) + δΓ (2) ,
where δ =1 if | ν t |≤ c and zero otherwise.
Several different threshold modeling procedures
have been developed. Here we utilize grid search
procedures to find the threshold value, c, that
minimizes the log of the determinant of the residual
covariance matrix, a procedure equivalent to
maximizing a normal likelihood function. We
constrain the grid search procedures to require each
regime to have at least twenty-five observations.
The parameters describing the two alternative
regimes are estimated conditional on the optimal
threshold values.
Once the parameters of the standard and regime
switching VAR models have been estimated,
standard methods of inference can be used to
evaluate the relationships among the prices and
protein variable. Here we utilize standard impulse
response functions to evaluate the dynamic
relationships among wheat class prices implied by
the alternative parameters. In threshold models,
several versions of the impulse responses could be
evaluated because in such models impulse
____________________________
4
This is widely recognized by the milling industry.
In the September 2003 International Trade
Commission (ITC) antidumping hearings with respect
to Canadian dumping of hard red spring wheat and
durum wheat, in oral testimony before the ITC U.S.
milling industry executives indicated that they tended
to determine blends of different wheat at the
beginning of each marketing year just after harvest
once the quality characteristics of different wheat
classes were known. Thereafter, they were generally
reluctant to change those blends.
responses may not be unique for alternative
observations or sizes of shocks. Potter’s nonlinear
impulse response analysis procedures could be used
to evaluate the responses at a particular observation
and allow for switching among regimes over the
period of the response.
Alternatively, impulses could be calculated at every
observation and mean responses or some other
summary measure then reported. Finally, impulse
responses could be evaluated at each alternative
regime with no shifting between regimes allowed
during the response. Here we adopt the latter
approach in that it yields the clearest inferences
regarding the differences in regimes.
Data and Empirical Results
We use monthly averages of daily cash prices for
three alternative classes of wheat—DNS in
Minneapolis, HRW in Kansas City, and SRW in
Chicago. The price data were collected from the
Bridge database. Average protein content for all
classes of U.S. wheat (HRW, DNS, SRW, durum,
and white wheats) for each crop year were provided
by annually published U.S. Wheat Associates Grain
Quality Reports. Quarterly stocks data were
obtained from unpublished NASS data.
We calculated an aggregate weighted average
protein content for the aggregate U.S. wheat harvest
each crop year using USDA statistics on production
for each class in each year to form weights. The
quarterly stocks data were multiplied by the protein
content of the crop to obtain “protein stocks” for
each quarter of the year. We then regressed this
protein stocks variable on the terms of a third order
Fourier series expansion. The data cover the 19892003 crop years.
The implied pattern of seasonality in protein is
illustrated in Figure 1. Note the presence of a large
increase in stocks with the influx of the winter
wheat harvest in June and July and then a second
smaller increase that occurs with the spring wheat
harvest in the late fall. Deviations from normal
protein levels are given by the deviations from the
seasonal patterns presented in Figure 1. We then
utilize cubic spline smoothing to interpolate the
quarterly protein stock measures to obtain monthly
observations. Such interpolation is most likely to
adequately represent data at a higher frequency in
cases where movements in the variable between
observations are likely to be smooth and gradual.
This is certainly the case for a highly aggregated
variable such as the total protein stocks implied for
the aggregate U.S. market. The observed and
interpolated protein stocks series are illustrated in
Figure 2. The blocks represent observed data while
the line represents the interpolated data used to
covert from quarterly to monthly frequencies.
Table 1 presents parameter estimates for a standard
VAR model. Parameter estimates for nonstructural
models of this form are usually of limited interest
and inferences are more efficiently extracted from
impulse responses. However, the coefficients on
the protein stocks variable are certainly of interest
in their own right. The coefficients are negative in
every case, suggesting that above-normal stocks of
protein are likely to have a depressing effect on
prices for each class of wheat. The coefficient is
largest in the case of the Kansas City hard red
winter price. The negative effect is also large for
the Minneapolis hard red spring price. The effect
for soft wheat prices in Chicago is much smaller
and not statistically significant.
These results are consistent with a priori
expectations. They imply that positive shocks to
the aggregate protein content of wheat in the U.S.
market have negative effects on hard red winter and
hard red spring wheat prices—high protein wheats
generally directed to uses demanding a high gluten
content. In contrast, the effect is not statistically
significant in the case of soft red winter wheat in
Chicago. Soft wheats are typically much lower in
protein content and are directed toward uses that
call for lower gluten wheats (e.g., cakes and
crackers rather than bread).
Impulse responses for the standard VAR model are
presented in Figures 3-6. Figure 3 illustrates the
dynamic paths of adjustment in prices to a positive
one-unit shock to the protein stocks variable. The
largest impact is realized by the Kansas City
price—a result entirely consistent with a simple
consideration of the VAR model protein
coefficients reported in Table 1. The impulse
indicates that a one unit increase in protein
generates a response of a 42 cent decrease in the
Kansas City per bushel price and a 34 cent decrease
in the Minneapolis per bushel price. In contrast, the
soft wheat price in Chicago shows only a small
negative response to the same protein shock. In
every case, the largest response occurs two months
after the shock, and that responses take ten or more
months to die out. This suggests that end users are
likely to be somewhat slow to adjust to protein
shocks and that market effects from such shocks
persist for several months. This finding seems to be
consistent with statements by U.S. millers at the
2003 ITC hearings on CWB dumping that they tend
to determine blends of different wheats for milling
on an annual marketing year basis after harvest and
to be relatively unresponsive to price changes.
normal protein levels are large. Again, the largest
effects are implied for Kansas City (hard red winter)
wheat prices. Large responses are also implied for
Minneapolis prices, although the adjustments are
somewhat smaller than those implied for the hard
red winter prices. This is not surprising because the
quantity of hard red winter wheat produced in the
U.S. is usually about twice as large as the quantity
of hard red spring wheat and hard red winter wheat
is therefore a more prominent source of aggregate
protein.
Adjustments to price shocks are modest once
protein shocks are accounted for. Minneapolis and
Kansas City prices appear to be more closely linked
that either market is with Chicago. The results
appear to imply a price leadership role for the
Minneapolis market in the Kansas City-Minneapolis
relationships. An innovation in the Kansas City
price results in almost no impulse response in the
Minneapolis price while an innovation the
Minneapolis price results in a smaller adjustment in
the same direction in the Kansas City price that
peters out after about 6 months.
Impulse responses for the alternative regimes are
presented in Figures 7-11. Figures 7 and 8, which
illustrate price responses to protein shocks in the
alternative regimes, are especially striking. A much
large response to a one unit shock to protein is
implied by the outside regime parameters. When
deviations from normal protein levels are more
modest, prices scarcely react at all. However,
significant adjustments occur when deviations from
normal protein levels are large. This result is
consistent with our hypothesis that large changes in
protein may have more significant effects on prices
than when protein shocks are small.5
As we have noted, price adjustment patterns may
reflect adjustment costs associated with changes in
production technologies that may be needed to
respond to substantial changes in wheat protein
availability. Table 2 reports estimates from a
threshold VAR model that allows shifting between
regimes according to the absolute value of the size
of shocks to the overall protein stocks available in
the market. The optimal threshold has a value of
0.1659. The band implied by this threshold that
defines alternative regimes is illustrated in Figure 2.
As one would expect, switching among regimes is
infrequent, reflecting the fact that the overall
availability of protein in the market is a slowly
adjusting variable. This implies that the market
tends to remain in a regime for an extended period
of time rather than jumping back and forth between
alternative regimes on a month to month basis.
Protein stocks generally have larger (more negative)
effects on prices in the “outside” regime, which
corresponds to periods when there are large
deviations from normal levels of protein. This is to
be expected in that, to the extent that costly
adjustments underlie the price relationships, such
adjustments are more likely to be undertaken and
are likely to be more extreme when deviations from
In the threshold VAR models, price adjustments are
similar to those found for the standard VAR model,
though again a much larger degree of price
responsiveness is implied in the outside regime.
This suggests that wheat prices are more responsive
to shocks in other markets when protein content is
above-normal or below-normal. However, the
impulse responses to do imply that when the price
of one class of wheat is shocked the prices of other
wheat classes adjust in very similar ways. This
suggests that wheat is not just wheat and that soft
red winter is by no means a perfect substitute for
hard red winter or hard red spring. Similarly, the
threshold model results also suggest that cross
market linkages between hard red spring wheat and
hard red winter wheat are complex.
____________________________
5
Note that our terminology may be somewhat confusing here.
All impulse response diagrams illustrate responses to
equivalent one-unit shocks. However, the regimes are defined
by the size of the protein shock. We could have presented
shocks that differed in terms of the size of the shocks in
alternative regimes. In such a case, the differences in impulse
responses would be exaggerated. Comparing the impulses at a
common level of shock allows a clearer view of how the
underlying structures of the models differ across regimes.
Conclusion
This study has utilized new data and innovative
econometric techniques to address a longstanding
issue in discussions about wheat markets - the
dynamic relationship between the prices of different
classes of wheat. A key data innovation consists of
the development and utilization of a measure of the
aggregate stock of protein in the U.S. wheat crop.
Data on average protein content by class of wheat
were combined with USDA statistics on production
by class and quarterly stock data to obtain protein
stocks for each quarter of the year. A third order
Fourier expansion was then utilized to obtain
estimates of normal protein levels that accounted
for quarterly seasonal effects. The quarterly data
were then interpolated using cubic splines to obtain
month-by-month estimates of protein stocks.
A key econometric and modeling innovation with
respect to wheat price dynamics has been the
utilization of a threshold modification of the VAR
model to account for potential adjustment costs
associated with changing use patterns of different
classes of wheat. The results from the estimated
threshold variant of the VAR model were also
compared with those from a standard VAR model in
which adjustment costs are ignored.
The major findings of the research are as follows.
In the standard VAR model, a positive one unit
shock to protein stocks has the largest and
statistically significant negative effect on the
Kansas City hard red winter price and a smaller but
still substantial effect on the Minneapolis hard red
spring price, as measured by impulse responses.
The impulse response of the Chicago soft red price
was not statistically significant and small.
Similar effects were identified in the threshold
model in which two regimes were identified: the
“inside” regime and the “outside” regime. The
“inside” regime is one in which protein levels do
not deviate very much in absolute terms (either up
or down) from normal seasonal levels. The
“outside” regime is one in which protein levels do
deviate substantially. The range within which
protein levels were deemed to be normal was
computed in the econometric estimation procedure.
In the threshold models, the effects of a unit change
in the protein stock level were qualitatively similar
to those reported for the standard VAR model.
When the absolute deviation of protein levels was
small (the “inside” regime), price impulse responses
were also small, and when the absolute deviation of
protein levels was large (the “outside” regime) the
impulse responses were much larger. In the outside
regime, the impulse response of the Kansas City
hard red winter price was much larger than the
impulse response of the Minneapolis hard red
spring price. These results are consistent with the
hypothesis that adjustment costs associated with
buyers (millers, etc) of different wheats changing
their patterns of use are relatively large.
In the threshold models, the effects of price shocks
for a specific class of wheat also depend on the
regime. However, one interesting result is that
shocks to the Kansas City hard red winter price
result in almost no impulse responses on the part of
Minneapolis Hard red spring prices, although
shocks to the Minneapolis price do generate a
qualitatively similar but smaller impulse response
on the part of the Kansas City price. In addition, an
exogenous shock to the Chicago soft red price
generates very weak impulse responses in the
Minneapolis price, although somewhat stronger
impulse responses in the Kansas City price.
These results also provide some further insights
about a long-standing argument between the
Canadian Wheat Board and U.S. wheat producers.
Fairly consistently, in a variety of wheat trade cases
brought before the U.S. ITC between 1992 and
2004, the CWB and its expert witnesses have
claimed that wheat is just wheat and, in particular,
hard red winter and hard red spring are almost
perfect substitutes for one another. The evidence
from this study tends to suggest that such is not the
case. The markets may be related but an exogenous
shock in the price of hard red winter wheat simply
does not generally result in a similar impulse
response in the price of hard spring wheat.
Table 1. Standard VAR Model of Wheat Prices: Parameter Estimates
Dependent
Variable
Chicago Price
Kansas City Price
Minneapolis Price
Explanatory
Variable
Constant
Protein Stocks (t)
Chicago Price (t-1)
Kansas City Price (t-1)
Minneapolis Price (t-1)
Chicago Price (t-2)
Kansas City Price (t-2)
Minneapolis Price (t-2)
Constant
Protein Stocks (t)
Chicago Price (t-1)
Kansas City Price (t-1)
Minneapolis Price (t-1)
Chicago Price (t-2)
Kansas City Price (t-2)
Minneapolis Price (t-2)
Constant
Protein Stocks (t)
Chicago Price (t-1)
Kansas City Price (t-1)
Minneapolis Price (t-1)
Chicago Price (t-2)
Kansas City Price (t-2)
Minneapolis Price (t-2)
Parameter
Estimate
37.3687
-20.7671
0.7734
0.2567
-0.0638
0.0341
-0.1826
0.0585
67.6815
-41.1863
-0.0504
1.0444
0.0112
0.0891
-0.2491
-0.0217
65.5465
-33.5678
-0.2288
0.5022
0.6614
0.2000
-0.4948
0.1993
Standard
Error
17.1231
13.1919
0.1183
0.1234
0.0922
0.1184
0.1220
0.0912
17.8796
13.7747
0.1235
0.1288
0.0962
0.1237
0.1274
0.0952
19.3841
14.9338
0.1339
0.1397
0.1043
0.1341
0.1381
0.1032
t
Ratio
2.18
-1.57
6.54
2.08
-0.69
0.29
-1.50
0.64
3.79
-2.99
-0.41
8.11
0.12
0.72
-1.96
-0.23
3.38
-2.25
-1.71
3.60
6.34
1.49
-3.58
1.93
Table 2. Threshold Switching Regime Model Parameter Estimates
Dependent
Variable
Chicago Price
Kansas City Price
Minneapolis Price
Explanatory
Variable
Constant
Protein Stocks (t)
Chicago Price (t-1)
Kansas City Price (t-1)
Minneapolis Price (t-1)
Chicago Price (t-2)
Kansas City Price (t-2)
Minneapolis Price (t-2)
Constant
Protein Stocks (t)
Chicago Price (t-1)
Kansas City Price (t-1)
Minneapolis Price (t-1)
Chicago Price (t-2)
Kansas City Price (t-2)
Minneapolis Price (t-2)
Constant
Protein Stocks (t)
Chicago Price (t-1)
Kansas City Price (t-1)
Minneapolis Price (t-1)
Chicago Price (t-2)
Kansas City Price (t-2)
Minneapolis Price (t-2)
Threshold Parameter
Proportion of Observations
Outside Regime
Parameter
Standard
Estimate
Error
22.7792
19.0585
-11.4098
14.6109
0.5035
0.1448
0.6560
0.1614
-0.1935
0.1357
0.2188
0.1459
-0.4298
0.1633
0.1451
0.1309
58.2909
19.5316
-37.0815
14.9736
-0.2723
0.1484
1.3526
0.1654
-0.0438
0.1390
0.2971
0.1496
-0.5197
0.1673
0.0367
0.1341
57.6586
21.6677
-24.4373
16.6112
-0.4385
0.1646
0.8157
0.1835
0.6917
0.1542
0.3664
0.1659
-0.6722
0.1856
0.0958
0.1488
0.1659
0.6278
t
Ratio
1.20
-0.78
3.48
4.07
-1.43
1.50
-2.63
1.11
2.98
-2.48
-1.84
8.18
-0.31
1.99
-3.11
0.27
2.66
-1.47
-2.66
4.45
4.48
2.21
-3.62
0.64
Inside Regime
Parameter
Standard
Estimate
Error
85.9199
36.7720
-13.1946
37.1174
1.3042
0.2093
-0.3652
0.2183
-0.0265
0.1213
-0.2871
0.1926
0.1911
0.1905
-0.0358
0.1229
144.7974
37.6848
-20.1873
38.0387
0.6244
0.2145
0.1662
0.2237
0.0378
0.1243
-0.0868
0.1974
-0.0640
0.1952
-0.0371
0.1260
106.6062
41.8062
-22.7985
42.1989
0.2421
0.2380
-0.1928
0.2482
0.5412
0.1379
0.0131
0.2190
-0.2158
0.2166
0.3566
0.1398
0.3722
t
Ratio
2.34
-0.36
6.23
-1.67
-0.22
-1.49
1.00
-0.29
3.84
-0.53
2.91
0.74
0.30
-0.44
-0.33
-0.29
2.55
-0.54
1.02
-0.78
3.92
0.06
-1.00
2.55
Figure 1. Estimated Seasonality in Protein Stocks Variable
Seasonality in Stocks*Protein
3
2.5
2
1.5
1
0.5
0
1
3
5
7
9
11
Month
Figure 2. Actual and Interpolated Protein Stocks Variable
Actual and Interpolated Protein Stocks
0.6
0.4
0.2
-0.6
Date
Jan-04
Jan-03
Jan-02
Jan-01
Jan-00
Jan-99
Jan-98
Jan-97
Jan-96
Jan-95
Jan-94
Jan-93
-0.4
Jan-92
-0.2
Jan-91
0
Jan-90
Deseasonalized Protein
0.8
Figure 3. Standard Impulse Responses to Protein Shocks
Impulse Responses to Protein/Stocks Shock
0
0
5
10
Cents/Bushel
-10
Chicago (srw)
Kansas City (hrw)
Minneapolis (hrs)
-20
-30
-40
-50
Periods After Shock
Figure 4. Standard Impulse Responses to Chicago Price Shocks
Impulse Responses to Chicago (srw)
Price Shock
Cents/Bushel
1.4
0.9
Chicago (srw)
Kansas City (hrw)
Minneapolis (hrs)
0.4
-0.1
1
2
3
4
5
6
7
8 9 10 11 12
-0.6
Periods After Shock
Figure 5. Standard Impulse Responses to Kansas City Price Shocks
Impulse Responses to Kansas City
(hrw) Price Shock
Cents/Bushel
1.4
0.9
Chicago (srw)
Kansas City (hrw)
Minneapolis (hrs)
0.4
-0.1
1
2
3
4
5
6
7
8
9 10 11 12
-0.6
Periods After Shock
Figure 6. Standard Impulse Responses to Minneapolis Price Shocks
Impulse Responses to Minneapolis
(hrs) Price Shock
Cents/Bushel
1.4
0.9
Chicago (srw)
Kansas City (hrw)
Minneapolis (hrs)
0.4
-0.1
1
2
3
4
5
6
7
8
9 10 11 12
-0.6
Periods After Shock
Figure 7. Outside Regime Impulse Responses to Protein Shock
Impulse Responses to Protein/Stocks Shock
0
1
3
5
7
9
11
13
Cents/Bushel
-10
-20
Chicago (srw)
Kansas City (hrw)
Minneapolis (hrs)
-30
-40
-50
Periods After Shock
Figure 8. Inside Regime Impulse Responses to Protein Shock
Impulse Responses to Protein/Stocks Shock
Cents/Bushel
12
8
6
4
10
-10
2
0
0
Chicago (srw)
Kansas City (hrw)
Minneapolis (hrs)
-20
-30
-40
-50
Periods After Shock
Figure 9.A. Outside Regime Price Responses to Chicago Price Shocks
Figure 10.A. Outside Impulse Responses to Kansas City Price Shocks
Impulse Responses to Chicago (srw)
Price Shock
Impulse Responses to Kansas City (hrw)
Price Shock
1.4
0.9
Chicago (srw)
Kansas City (hrw)
Minneapolis (hrs)
0.4
-0.1
1
2
3
4
5
6
7
8
Cents/Bushel
Cents/Bushel
1.4
9 10 11 12
-0.6
0.9
Chicago (srw)
Kansas City (hrw)
Minneapolis (hrs)
0.4
-0.1
1
2
3
4
5
6
7
8
9 10 11 12
-0.6
Periods After Shock
Periods After Shock
Figure 9.B. Inside Regime Price Responses to Chicago Price Shocks
Figure 10.A. Inside Impulse Responses to Kansas City Price Shocks
Impulse Responses to Chicago (srw)
Price Shock
Impulse Responses to Kansas City
(hrw) Price Shock
1.4
0.9
Chicago (srw)
Kansas City (hrw)
Minneapolis (hrs)
0.4
-0.1
1
2
3
4
5
6
7
8
9 10 11 12
-0.6
Cents/Bushel
Cents/Bushel
1.4
0.9
Chicago (srw)
Kansas City (hrw)
Minneapolis (hrs)
0.4
-0.1
1
2
3
4
5
6
7
8
9 10 11 12
-0.6
Periods After Shock
Periods After Shock
Figure 11.A. Outside Impulse Response to Minneapolis Price Shocks
Impulse Responses to Minneapolis
(hrs) Price Shock
Cents/Bushel
1.4
0.9
Chicago (srw)
Kansas City (hrw)
Minneapolis (hrs)
0.4
-0.1
1
2
3
4
5
6
7
8
9 10 11 12
-0.6
Periods After Shock
Figure 11.B. Inside Impulse Response to Minneapolis Price Shocks
Impulse Responses to Minneapolis (hrs)
Price Shock
1.6
Cents/Bushel
1.1
Chicago (srw)
Kansas City (hrw)
Minneapolis (hrs)
0.6
0.1
-0.4
1
2
3
4
5
6
7
8
9 10 11 12
-0.9
Periods After Shock
References
J. N. Barnes and D. A. Shields, “The Growth in U.S. Wheat Food Demand,” Wheat Yearbook, 1998, USDA
Economic Research Service, pp 21-29
Espinosa, J. A., and B. K. Goodwin. “Hedonic Price Estimatoin for Kansas Wheat Characteristics.” Western
Journal of Agricultural Economics 16 (1991): 72-85.
Goodwin, B. K., and T. Schroeder. “Price Dynamics in International Wheat Markets,” Canadian Journal of
Agricultural Economics, 39(1991): 237-54.
T. L. Marsh, “Elasticities for U.S. Wheat Food by Class,” unpublished Working Paper, Kansas State University,
September 2002.
Mohanty, S., W. H. Meyers, and D. B. Smith. “A Reexamination of Price Dynamics in the International Wheat
Market,” Canadian Journal of Agricultural Economics v47 (1999): 21-29.
Parcell, J. L., and K. Stiegert. “Competition for U.S. Hard Wheat Characteristics,” Journal of Agricultural and
Resource Economics, 23 (1998): 140-154.
Potter, S. M. “A Nonlinear Approach to US GNP," Journal of Applied Econometrics. 10(April-June 1995):109
125.
Stiegert, K, and J-P. Blanc. “Japanese Demand for Wheat Protein Quantity and Quality,” Journal of
Agricultural and Resource Economics, 22 (1997): 104-119.
United States International Trade Commission. Wheat, Wheat Flour, and Semolina Investigation 22-5.
Publication 2794, Washington ,D.C., July 1994.
Copyright 2005:
The programs of the MSU Extension Service are available to all people regardless of race, color, national origin, gender, religion, age, disability,
political beliefs, sexual orientation, and marital or family status. Issued in furtherance of cooperative extension work in agriculture and home
economics, acts of May 8 and June 30, 1914, in cooperation with the U.S. Department of Agriculture, Dr. Douglas Steele, Vice Provost and Director,
Extension Service, Montana State University, Bozeman, MT 59717.